Procedural Steps for Knowledge Mining in Time Series

Abstract

intervals which form temporal patterns are usually formulated through Allen's interval relations that originate in temporal reasoning. But this representation is not advantages for knowledge discovery. The Hierarchical Time series Knowledge Representation (HTKR) is the hierarchical language which expresses the temporal aspects of coincidence and partial order, for interval patterns. We present mining procedural steps which are more efficient, effective and based on item set techniques. Pruning of the search space minimizes the mining result size considerably, thereby speeding up the procedural steps and easing the interpretations. When applied on the real data set, HTKR can provide the explanation of underlying temporal phenomena, but whereas the numerous Allen's relation patterns only explains fragmented data.